fix: llama4 vllm agg multimodal script (#6103)

docs/features/multimodal/multimodal_vllm.md

@@ -224,7 +224,7 @@ flowchart LR
 
 ```bash
 cd $DYNAMO_HOME/examples/backends/vllm
-bash launch/agg_multimodal_llama.sh
+bash launch/agg_multimodal.sh --model meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8
 ```
 
 **Client:**

examples/backends/vllm/launch/agg_multimodal.sh

@@ -58,6 +58,8 @@ if [[ "$MODEL_NAME" == "Qwen/Qwen2.5-VL-7B-Instruct" ]]; then
     MODEL_SPECIFIC_ARGS="--gpu-memory-utilization 0.85 --max-model-len 4096"
 elif [[ "$MODEL_NAME" == "llava-hf/llava-1.5-7b-hf" ]]; then
     MODEL_SPECIFIC_ARGS="--gpu-memory-utilization 0.85 --max-model-len 4096"
+elif [[ "$MODEL_NAME" == "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8" ]]; then
+    MODEL_SPECIFIC_ARGS="--tensor-parallel-size=8 --gpu-memory-utilization 0.85 --max-model-len=108960"
 fi
 
 # Start vLLM worker with vision model

examples/backends/vllm/launch/agg_multimodal_llama.sh

-#!/bin/bash
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
-# SPDX-License-Identifier: Apache-2.0
-set -ex
-
-trap 'echo Cleaning up...; kill 0' EXIT
-
-MODEL_NAME="meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
-
-# run ingress
-# dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
-python -m dynamo.frontend &
-
-# run processor
-python -m dynamo.vllm --multimodal-processor --enable-multimodal --model $MODEL_NAME &
-# Llama 4 doesn't support image embedding input, so use encode+prefill worker
-# that handles image encoding inline
-python -m dynamo.vllm --multimodal-encode-prefill-worker --enable-multimodal --model $MODEL_NAME --tensor-parallel-size=8 --max-model-len=208960 --gpu-memory-utilization 0.80 &
-
-# Wait for all background processes to complete
-wait

fern/pages/features/multimodal/multimodal-vllm.md

@@ -212,7 +212,7 @@ flowchart LR
 
 ```bash
 cd $DYNAMO_HOME/examples/backends/vllm
-bash launch/agg_multimodal_llama.sh
+bash launch/agg_multimodal.sh --model meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8
 ```
 
 **Client:**